Insects are precious. However, they are doing badly. For several years, dozens of scientific studies have warned of the worrying decline in their populations. This is also one of the central subjects of COP16 on biodiversity, which has been taking place since Monday October 21 in Colombia. However, the number of insect species – more than a million – and still incomplete data do not allow us to precisely quantify the extent of the damage. And recently, some work has even suggested that the situation is not that catastrophic.
This is the case of a study published in 2020 in the prestigious scientific journal Science. This meta-analysis – one of the highest levels of scientific evidence – brings together the results of 166 studies and compiles records from 1,676 insect counting sites around the world. This work received significant media coverage which helped to disseminate a reassuring message to the general public. Enough to breathe a “phew” of relief? Not for dozens of doubtful specialists who studied meta-analysis from every angle, before discovering numerous irregularities. They then reported these issues to authors of the meta-analysis and to the review Science. But only a minimalist erratum was published, leaving the results of the study unchanged.
A “work of ants”
The story could have ended there. But a year later, the authors of the meta-analysis published In Ecologyone of the most renowned journals in the field of ecology, their immense “InsectChange” database on which they based their work and which compiles all the insect records used. A perfect opportunity to precisely verify the results of the meta-analysis for those who had the courage to take the plunge.
This was the case of two French women, Marion Desquilbet, economist at the Toulouse School of Economics and at INRAE, and Laurence Gaume, ecologist at the CNRS and the University of Montpellier. After months of work, they published the results of their investigations in Peer Community Journa magazineL and reveal that the InsectChange database contains no fewer than 553 errors. And not least. “We have done real painstaking work to dissect everything and have uncovered a multitude of new problems whose origin and nature we are deciphering,” explains Laurence Gaume. Nothing can resist the relentlessness of an ant.
To fully understand the reasons for these errors, we must remember how we count insects. “There are several ways according to studies: by visual counting by following a route on which observers count the butterflies; with web traps to collect flying insects, pits for crawling insects, aquatic nets, etc.” , details Marion Desquilbet. Once the insects are cataloged, researchers write down the numbers in data sets that they fill out at different times of the year, with monitoring sometimes going back several decades. But these files can contain very diverse information, such as the number or weight of insects, or include vertebrate or non-vertebrate animals, larger or smaller geographical areas, etc.
Invasive molds counted as “insects”
It is therefore not enough to collect all the data and compile it. It is also necessary to check their quality and representativeness. A complex job that the authors of the meta-analysis obviously did a little too quickly. Incorrect counts when transmitted from one database to another, sampling bias, non-standardized units of measurement, inadequate geographic coordinates of sampling… The list of problems is long.
The two researchers discovered, for example, that datasets concerning freshwater environments do not only include insects, but also aquatic invertebrates. The authors of the meta-analysis thus reached the conclusion that the biomass – the weight of insects – is increasing in certain aquatic environments even though they included a species of invasive mussels (which threaten the biodiversity of their environment) in their calculations. . Not only are they not insects, but mussels weigh much more than the majority of aquatic insects, which has doubly skewed the calculations. According to Marion Desquilbet and Laurence Gaume, 40% of abundance data [NDLR : le nombre d’insectes] and 80% of InsectChange biomass data contains non-insects that have contributed to falsely increasing insect populations.
InsectChange also takes into account surveys of dragonfly populations which come from experimental ponds created by an English scientist who wanted to study their colonization of these spaces. “These data are incorporated without mention of the experimental context and cause an – artificial – increase in the dragonfly population,” lament the two researchers. In total, the researchers calculated that the database includes five times more contexts favorable to the increase of insects than to their decrease.
The authors of InsectChange also wanted to compare the different insect population surveys to satellite data in order to understand the impact of land use on the evolution of these insects. But again, they made big mistakes. Overall, InsectChange has inaccurate geographic coordinates in two-thirds of the datasets. This methodology led to errors in estimating the impact of agricultural crops, urbanization and global warming on insect declines. “This shows a total absence of methodological reflection regarding the assembly of databases,” notes Marion Desquilbet. Questioned by L’Express, the authors of the meta-analysis and the InsectChange database did not respond to our requests.
The decline of insects, the most likely hypothesis
The work of the two researchers is all the more important as other scientific studies have relied on InsectChange to work. Unsurprisingly, their results are also distorted. The investigation by French researchers recalls a sad reality: insects are indeed in danger and actions to preserve them must be continued. “The meta-analysis of Science indicates that the decline in insect populations is only 9% per decade. This is already not nothing, but it remains very much lower than all the other studies, points out Marion Desquilbet. We have been observing this decline for forty years and it is accelerating. There is a fire, we must no longer wait to have a thermometer that tells us the temperature, we must use water lances now.”
Because the Earth has an incredible number of diverse and extremely connected animal species. In this natural house of cards, insects constitute an essential piece. Their decline can lead to a cascade of harmful events for all living species, both plants and animals – because insects are pollinators, but also food reservoirs for fish, birds, etc. – as well as for human economies.
Science also in danger?
In addition to the decline of insects, Marion Desquilbet and Laurence Gaume are worried about science. They believe that journals should in particular take better account of the criticisms formulated by the scientific community after the publication of a study. “It is even more important for prestigious magazines which are the privileged intermediaries of journalists,” they emphasize. It may indeed appear surprising that Science did not investigate further after the alert given by dozens of researchers. “This raises the question of the economic model of scientific journals, which are encouraged to publish surprising results to get people talking about them in order to improve their impact factor [NDLR : indicateur estimant la visibilité d’une revue]”, they indicate.
“Our study shows that it is essential to ensure the quality of global databases, whether in ecology or in other fields, because if we do not ask this question, where is science going?” insists Laurence Gaume again. According to the two scientists, there is a severe shortage of specialists in database analysis. They therefore added, in their study, a reproducible analysis method. Enough to inspire the development of database quality evaluation grids.
.